F.-Y. Dupradeau
---
http://q4md-forcefieldtools.org/FyD/
attached mail follows:
Dear Michael,
Considering ethane & n-alkanes more generally:
Please, see
http://onlinelibrary.wiley.com/doi/10.1002/jcc.540150705/abstract
The problem you report is related to rounding off errors: Rounding off
errors are really a problem... If the accuracy requested in the
program output is at 6 digits as in the RESP output, the max. error
that is accumulated is +/-Y.00000X (Y is an integer & X is around 6-7
from our tests). But, here, who cares ;-) as X.10-6 remains a small
value...
Then, if the accuracy requested is at 4 digits (Cornell et al. force
field) the rounding off errors becomes +/-Y.000X, which remains
acceptable; this becomes the pain if the accuracy requested is at 3
digits (GLYCAM) or even 2 digits (CHARMM_?).
Then, an additional problem is that when correcting these rounding off
errors, chemical equivalencing should not be broken; at 10-6, it does
not really matter once again I guess: but what about for an accuracy
of 10-3? We have developed an algo. related to this rounding off
errors in R.E.D. We should released it in soon in R.E.D. Server first.
The idea of this algo. is to automatically correct rounding off errors
in the context of a force field topology database generation (large
set of FF libraries) in agreement with (i) the charge accuracy
requested by the user, (ii) chemical equivalencing required in MD
simulation, and (iii) the constraint values requested during the fit.
regards, Francois
> I wonder if there is something wrong with either my approach or with the
> antechamber process.
> Something as simple as ethane gives charges that do not add to zero
> (+0.00200).
> input pdb attached and output mol2 file.
> This is the command:
>
> antechamber -i ethane.pdb -fi pdb -o ethane.mol2 -fo mol2 -c bcc -s 2
>
> Motivation:
> We use Amber, Namd, and charmm a lot imoving between programs for
> different purposes on the same system and it is pretty important that each
> program is getting the forces right for each force field.
> In an effort to test force fields in each program (we are also testing
> gromacs and LAMMPS, though we donšt use them much), we are making some
> simple test cases.
> This test is meant to test amber ff (gaff in this case) in the other
> programs and compare results to amber.
> We create simple molecules in the amber process and then run them in the
> other programs converting input as needed.
> Problem starts with antechamber, giving a non-neutral alkane.
> If I manipulate the charges to be neutral, we can continue, but I think
> antechamber should do that.
> Am I expecting too much? Is the collective wisdom on antechamber that it
> is imprecise enough that manipulating the charges is no worse than what
> you are getting anyway?
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Mon Jan 16 2017 - 22:00:02 PST
